What will you be able to do?

By the end of this module, you will be able to run a PyTorch-to-ExecuTorch deployment workflow for transformer inference on an Arm CPU.

You will be able to:

  • prepare GPT-2 and DistilBERT for fixed-shape export

  • lower exported PyTorch programs into portable and XNNPACK-backed ExecuTorch .pte artifacts

  • inspect delegated and fallback graph regions with Model Explorer

  • benchmark PyTorch eager, portable ExecuTorch, and ExecuTorch with XNNPACK on a Raspberry Pi

  • run small application demos using delegated .pte artifacts

  • compare artifact size, latency, delegation behavior, and application fit

  • choose an appropriate next deployment path for an Arm CPU target

  • identify next steps for production deployment of transformer models

Previous page Complete and continue